Digital climate
Welcome!
This session, we look at the environmental impacts of the digital industries
Intro
Last time, we talked about artificial intelligence, and the rise of GPTs due to the increase in data and hardware.
Plan for the day
- Digital footprints
- A tale of two transitions
- Low-tech
- Group work
What is the environmental toll of a Netflix binge?
Software impact
A webpage is about 5,000 Kb.
An HD series episode is about 7,000,000 kb.
The first immediate problem with this the increase in the amount of data being transferred. Streaming in HD is more than streaming in SD. And we've been streaming more and more.
Streaming moved from adapting to bandwidth restriction, to bandwidth adapting to streaming's desires1.
In 2009, the IT sector surpasses air travel in CO2 emissions.
And it keeps growing.
Nathan Ensmenger argues the history of computing was always about extracting and managing resources, see Ensmenger, Nathan, The Environmental History of Computing, Technology and Culture, Volume 59, Number 4 Supplement, October 2018, https://homes.luddy.indiana.edu/nensmeng/files/Ensmenger2018.pdf
Data centers' energy requirement could double by 2026, represent 7-12% of total U.S. domestic consumption2
The report from the DoE referenced mentions: "The report finds that data centers consumed about 4.4% of total U.S. electricity in 2023 and are expected to consume approximately 6.7 to 12% of total U.S. electricity by 2028. The report indicates that total data center electricity usage climbed from 58 TWh in 2014 to 176 TWh in 2023 and estimates an increase between 325 to 580 TWh by 2028.
For instance, Microsoft has re-opened a nuclear powerplant reactor to power its AI needs: https://www.npr.org/2024/09/20/nx-s1-5120581/three-mile-island-nuclear-power-plant-microsoft-ai
Lifecycle breakdown of the emissions34:
- device manufacturing (factories)
- data centers (storing content, running algorithms)
- network infrastructure (delivering content)
- user-side operations (battery use)
- e-waste management (throwing things away)
For more information on how cobalt and coltan are essential to batteries, and what that means for the lives of people in Congo, see https://www.npr.org/sections/goatsandsoda/2023/02/01/1152893248/red-cobalt-congo-drc-mining-siddharth-kara
Device manufacturing is mainly done in China (for now, highly coal-intensive, might change).
It needs water for:
- itself (cooling)
- its electricity source (power and cooling)
And what it takes, it prevents others from using. For instance, in Ireland, data centers 11% of the electricity grid, and that puts the rest of the country under strain.
Solutions to mitigate emissions include shifting to renewable energy, energy-efficient devices, circular economy principles, edge computing, green coding, and improved recycling practices. The ICT sector can significantly reduce its environmental impact and contribute to global sustainability efforts by adopting these solutions.
Beware of the rebound effect.
A tale of two transitions
Is the digital transition supporting the climate transition?
This includes geo-engineering, nuclear, carbon capture, iot, smart city, etc.
Most assumptions is that the energy transition will be supported by the digital transition; that through dematerialization and efficiency gains, we will be able to live life as usual without having to fundamentally change lifestyles. See for instance https://www.ecomodernism.org
Can we ever decouple from nature?
- geo-engineering
- nuclear
- carbon capture
- iot
- smart city
- etc.
This very modernist approach, and technno-optimistic approach might have its limits.
See Gauthier Roussilhe, Perspectives and Methodologies https://gauthierroussilhe.com/en/articles/nouvelles-perspectives-de-recherche
Changes of perception as we see the Earth through the digital56.
Historically, there have been no "transitions", energy and material have only accumulated7.
It is also necessary to look at which point of the system are we operating8.
Most of what we do in the lower part. But the higher part has the most leverage.
The goal of the system is profit.
One mindset change was "one apple iphone every year"
Meanwhile, Apple used to force recyclers to shred Macbooks and iPhones, rather than re-use them9.
Apple has managed to build up a mindset of "one iphone every year", or at least profits massively from it.
Even in the recycling process, half of the materials get lost due to the toxic chemicals used: https://www.ifixit.com/News/94386/the-truth-about-apples-free-iphone-recycling-program-the-earth-deserves-better
Repair and refurbishment should be supported, and smarter designs developed, to extend the lifetime of EEE. The easiest solution for all e-waste issues is still not to generate any e-waste in the first place.10
Landfill < e-waste recyclng < re-use < repair
Low-tech
A change of design, and a change of mindset.
The way forward is not more efficient, but less greedy.
Direct, individual servers sharing solar resources: website and servers .
Sometimes, it's OK for things to turn off.
Caring for things makes them tangible and precious11, and depends on a right to repair12.
Concretely, this translates into the adoption of Right to Repair directives (2024) in the EU.
In general, a low-tech approach:
- questions the intrinsic desirability of technological progress
- questions our needs (maybe they're just wants?)
- questions whether tech alternatives can exist without social alternatives
Group work
time to do group work
Next steps:
- make sure you have updated your URL on the document
- individual commentary due on 05/09, 18:00
- each group meets online, week of 08/09
- final due date 19/09, 18:00
Course wrap-up
Take 3 minutes and write down something you remembered from this week.
Technology is a kind of magic, but it's still made by humans.
Algorithms are only as dangerous as the social contexts they support.
There are more than a few consequences.
Possibility is not intrinsic desirability.
Change your passwords, reject cookies and use Sci-Hub.
Thanks!